MCP for dummies.
It's basically a USB-C connector.
Published Sep 4, 2025 · Around 2 minutes to read
I use one USB-C cable on my Macbook for charging, connecting my second monitor, and connecting to the internet. One cable, three functions.
MCPs are like a USB-C connector, but for AI. It’s a universal translator that allows AI work with YOUR data. Without MCP, AI is limited to giving you basic answers, because it’s trained on generic data. With MCP, you get your own personal assistant: it can read your files, fetch your calendar conflicts, or get the design tokens from your Figma file. One cable, multiple functions.
Another way to think about MCP is like a text input box. An empty ChatGPT input box invites to you put stuff in it. Text at first. But you eventually learn that it’s flexible enough to add images, videos, audio, spreadsheets etc. And the AI can make sense of it all.
MCP lets you personalize your AI.
—
P.S. I built an MCP server (to get the weather from a couple of cities) and wrote this post in an attempt to understand MCP better. I tested my tools in the MCP Inspector so I wouldn’t have to waste tokens. I hooked the server up to Claude Desktop.
P.P.S. The Figma MCP server is read-only. Which means I can only get design node information. It’s great for the dev part of the workflow, like creating design system components. But I wish I could programmatically edit Figma files, mostly for design variations.
READ NEXT